Cornell engineers have developed a powerful artificial intelligence tool that could help New York State and other governments plan for the transition to a carbon-neutral energy sector, using a combination of machine learning and modeling optimization to provide hour-by-hour analysis of the state of the empire. energy needs.
States, including New York, which has committed to producing 100% clean electricity by 2040, are using technological, environmental and economic data to determine the best policy and investment choices to integrate more renewable energy in the network. But from a computational perspective, the modeling challenge is huge, said Fengqi You, Roxanne E. and Michael J. Zak Professor of Energy Systems Engineering at Cornell Engineering.
“There are design decisions such as how many solar panels or wind turbines to install and how much energy storage capacity to build,” said You, principal investigator at the Cornell Atkinson Center for Sustainability, “but even more complex are the hourly operation decisions such as how much power goes from upstate to downstate, or from storage center to neighborhood.”
You said such high-resolution planning can be achieved using “multi-scale bottom-up optimization” modeling combined with machine learning. The framework is detailed in the Feb. 7 print edition of the journal ACS Sustainable Chemistry & Engineering. The study was co-authored by graduate student Ning Zhao.
The research builds on You’s 2019 study that showed how modeling can help guide New York’s long-term energy goals. But the modeling of annual energy supply and demand does not take into account peaks in demand that occur hour by hour. New York’s unstable weather causes wild fluctuations in electricity demand and intermittent power from sources such as wind and solar.
To illustrate their new energy transition framework, You and Zhao produced case studies on the decarbonization of New York’s electric power, optimizing annual capacity planning and hourly system operations, while incorporating data on the technology, capacity and age of power generation and storage facilities everywhere. the state.
“We’re trying to bring technologies like machine learning, data analytics, optimization, and artificial intelligence to help a state understand what’s needed to operate not just every year, but also every hour. with renewable energy,” You said.
In one case study, which proposed to increase electricity storage capacity in New York, the transition model indicated that the total electricity generation capacity was 39% higher than in another case without storage. extended. If the state chose not to expand electricity storage, it would require 200% more generation capacity based on non-intermittent power.
Detailed hourly simulations indicated that offshore wind, hydro and solar are the optimal energy sources by 2040, but if electricity storage capacity could not be increased tenfold, options for Solar energy should be replaced by nuclear in order to create a reliable source of energy. energy network.
“It’s exciting to see the whole transition process we’ve achieved with these optimization tools,” Zhao said. “This can provide a lot of insight into how our future system might look and how we can advance this transition to decarbonization in an economically efficient and reliable way.”
The research was supported by the National Science Foundation.
Syl Kacapyr is PR and content editor for the College of Engineering.